We could be in for a delightfully wild ride in graphics once Computex and E3 roll around, or perhaps shortly thereafter. AMD will kick things off with a new round of graphics cards based on its Navi GPU architecture, and those could be interesting in and of themselves. Where things could truly heat up, however, is with how NVIDIA responds, and rumor has it the GPU maker is prepping a refresh of its high-end Turing parts.
To this point, NVIDIA has focused on bringing Turing to the masses by following up the launch of pricey (and full featured) RTX parts with less expensive GTX models, including the GeForce GTX 1660 Ti and GeForce GTX 1660, and more recently the GeForce GTX 1650. All of those cards lack dedicated RT and Tensor cores for real-time ray tracing and Deep Learning Super Sampling (DLSS), but are otherwise based on NVIDIA's latest-generation Turing GPU architecture.
In a matter of weeks, AMD is expected to counter with its first-generation of Navi. This will not be the same version of Navi that ends up in Sony's PlayStation 5 console, and so it remains to be seen if the first-gen part models support ray tracing. Navi could, however, bring performance parity to some of NVIDIA's higher-end graphics cards, like the GeForce RTX 2070. Depending on how Navi parts are priced, AMD could have a compelling lineup on its hands.
This is where the latest rumor comes into play. According to YouTube channel RedTechGaming, NVIDIA is planning to launch a series Turing GPUs with upgraded 16Gbps GDDR6 memory, compared to the 14Gbps GDDR6 memory that it uses now. Looking strictly at the memory, here is a breakdown of NVIDIA's current RTX cards...
- GeForce RTX 2080 Ti: 11GB of GDDR6 at 14Gbps
- GeForce RTX 2080: 8GB GDDR6 at 14Gbps
- GeForce RTX 2070: 8GB GDDR6 at 14Gbps
- GeForce RTX 2060: 6GB GDDR6 at 14Gbps
Every RTX card uses GDDR6 memory clocked at 14Gbps, ranging from 6GB on the GeForce RTX 2060 to 11GB on the GeForce RTX 2080 Ti.
Things are a little different on NVIDIA's GTX cards based on Turing. Here's how those shake out...
- GeForce GTX 1660 Ti: 6GB GDDR6 at 12Gbps
- GeForce GTX 1660: 6GB GDDR5 at 8Gbps
- GeForce GTX 1650: 4GB GDDR5 at 8Gbps
As the rumor goes, only the RTX lineup would see a refresh, though it's unclear how exactly NVIDIA might approach that. Would it stick to the same model names, but sell newer versions with faster memory and potentially higher GPU clocks? Or would NVIDIA launch brand new model names, like a GeForce RTX 2070 Ti or a GeForce RTX 2090?
That is not known for sure, though it's worth pointing out that NVIDIA followed the former strategy in the past with Pascal—for example, the GeForce GTX 1080 launched 8GB of GDDR5X memory at 10Gbps, but was later bumped to 11Gbps.
It's all rumors and speculation at this point. Should NVIDIA decide to refresh its RTX lineup, we could see both faster memory and faster GPU clocks. NVIDIA could also (or alternatively) lower pricing on its RTX cards, all in an attempt to steal some thunder from AMD's Navi launch. We will have to wait and see.